SimHacker avatar

yaml-jazz

YAML is sheet music. The LLM is the jazz musician. Comments are soul.

作者 SimHacker|オープンソース

🎷 YAML Jazz

YAML is sheet music. The LLM is the jazz musician. Comments are soul.

MOOLLM K-Lines

K-LineWhy Related
moollm/Core MOOLLM principle
postel/Liberal interpretation of ambiguous YAML
mind-mirror/Personality as YAML Jazz (numbers + comments = life)
needs/Dynamic comments as character inner voice
sniffable-python/YAML Jazz applied to Python comments
empathic-templates/Semantic understanding of YAML
files-as-state/YAML IS the state format
k-lines/Comments activate conceptual clusters
character/Character state in YAML

Quick Links:

Overview

YAML Jazz is registered in schema (schemapedia) as mechanism yaml-jazz under family notation—a first-class schema plugin alongside interchange and causal entries. Pair with JSON Schema when you need validated tree plus comment-rich source.

YAML Jazz is how MOOLLM treats structured data: not as rigid schemas, but as semantic improvisation where:

  • Structure carries meaning — indentation, ordering, grouping
  • Comments ARE data — they're guidance, context, soul
  • The LLM interprets — filling gaps, resolving ambiguity
  • Schemas are suggestions — starting points, not prisons

Comments Matter

# CRITICAL: Do not modify without user approval
api_key: ${SECRET}

hunger: 3  # Getting peckish. Is that pie?
           # ← This comment IS the character's inner voice

The LLM reads these. Acts on them. Comments are instructions.

Core Principle

No decorative ---- or ==== line comments — they waste tokens!

This is a kernel-level directive.

Standards, dialects, and lenses

YAML Jazz sits on top of the YAML spec (1.1 / 1.2 in the wild) and whatever domain pack you use (Kubernetes, Actions, MOOLLM skills, …). The skill file is the source of truth for humans and LLMs; many parsers strip comments on load, so preserve the text or use round-trip tooling when machines rewrite files.

Comments use full-line # blocks for narrative and end-of-line # for field-local gloss. Both are semantic.

Documentation by example: the YAML instance teaches usage; comments add intent, guardrails, and counterexamples.

Parallel copies: the same key–value tree can appear in multiple files, each with a different lens (intent, validation, metrics, history). See examples/lens-intent.yml and examples/lens-validation.yml, and SKILL.md for the full protocol.